Management of malignant colonic polyps: A population-based analysis of colonoscopic polypectomy versus surgery


  • Gregory S. Cooper MD,

    Corresponding author
    1. Division of Gastroenterology, University Hospitals Case Medical Center, Cleveland, Ohio
    2. Case Comprehensive Cancer Center, Cleveland, Ohio
    • Division of Gastroenterology, University Hospitals Case Medical Center, 11100 Euclid Avenue, Cleveland, OH 44106-5066
    Search for more papers by this author
    • Fax: (216) 983-0347

  • Fang Xu MS,

    1. Division of Gastroenterology, University Hospitals Case Medical Center, Cleveland, Ohio
    2. Department of Epidemiology and Biostatistics, Case Western Reserve University, Cleveland, Ohio
    Search for more papers by this author
  • Jill S. Barnholtz Sloan PhD,

    1. Case Comprehensive Cancer Center, Cleveland, Ohio
    2. Department of Epidemiology and Biostatistics, Case Western Reserve University, Cleveland, Ohio
    Search for more papers by this author
  • Siran M. Koroukian PhD,

    1. Case Comprehensive Cancer Center, Cleveland, Ohio
    2. Department of Epidemiology and Biostatistics, Case Western Reserve University, Cleveland, Ohio
    Search for more papers by this author
  • Mark D. Schluchter PhD

    1. Case Comprehensive Cancer Center, Cleveland, Ohio
    2. Department of Epidemiology and Biostatistics, Case Western Reserve University, Cleveland, Ohio
    Search for more papers by this author

  • We acknowledge the efforts of the Applied Research Program, National Cancer Institute; the Office of Research, Development and Information, Centers for Medicare & Medicaid Services; Information Management Services, Inc.; and the Surveillance, Epidemiology, and End Results (SEER) Program tumor registries in the creation of the SEER-Medicare database.



The management of colon polyps containing invasive carcinoma includes surgical resection or colonoscopic polypectomy. To date, there are very limited population-based data comparing outcomes with the 2 management approaches.


Using the linked Surveillance Epidemiology and End Results–Medicare database, we identified 2077 patients aged ≥66 years with an initial diagnosis of stage T1N0M0 malignant polyp from 1992-2005. Patients were categorized as surgical or polypectomy depending on the most invasive treatment. To adjust for potential selection bias in treatment assignment, using multivariate analysis, patients were divided into quintiles of likelihood of polypectomy (propensity scores), and outcomes were compared in each quintile.


Surgical resection was performed in 1340 (64.5%) patients and polypectomy was performed in 737 (35.5%) patients. Predictors for undergoing polypectomy (P<.001) included older age, greater comorbidity, no history of polyps, diagnosis in 2002 or later, left colon site of cancer, well-differentiated tumors, and colonoscopy performed in an outpatient setting. Both 1-year and 5-year survival were higher in the surgical group (92% and 75%, respectively) than in the polypectomy group (88% and 62%, respectively). The unadjusted hazard ratio was 1.51 (95% confidence interval [CI], 1.31-1.74). After adjusting for propensity quintile, the hazard ratio was 1.15 (95% CI, 0.98-1.33). Within each propensity quintile, the risk of death was similar between the 2 groups (interaction test P = .96).


In this large, population-based sample, more than one-third of patients with malignant polyps were treated with colonoscopic polypectomy. Outcomes were similar to surgical patients with comparable clinical characteristics and could be offered to patients who meet appropriate clinical criteria. Cancer 2012;. © 2011 American Cancer Society.

A malignant polyp is an adenoma in which malignant cells have spread into the deeper substance of the polyp, including involvement of the stalk.1 At the time of colonoscopy, it is usually impossible to recognize an adenoma that contains invasive cancer or even one that is completely replaced by cancer, termed a polypoid carcinoma. Thus, these lesions are typically treated initially the same as benign polyps, with polypectomy. Once diagnosed pathologically, further management of malignant polyps is somewhat controversial, with the 2 main options being referral for surgical resection or observation after colonoscopic polypectomy. Pathological criteria that are thought to favor colonoscopic cure include complete excision from the endoscopist's perspective, well-differentiated tumors, lack of involvement of the cautery margin with cancer, lack of involvement of the polyp stalk with cancer, and lack of invasion of the venous or lymphatic channels.2-5 In addition, sessile or flat polyps are thought to have a higher risk of metastatic spread than pedunculated or polyps on a stalk. Nonetheless, even with application of these criteria, there is debate about the best approach.

Proponents of the surgical approach argue that it provides complete staging and that a certain proportion of patients with ostensibly localized disease are found to have regional nodal metastases. Moreover, surgery eliminates the likelihood of early local recurrence due to incomplete removal. Currently, many resections can be performed laparoscopically, which has been associated with improved postoperative recovery. The major advantage of the colonoscopic approach is that it avoids the potential morbidity and mortality of bowel resection. Whereas the 2 approaches have never been compared in a randomized controlled trial, several single-institution case series have evaluated patient outcomes with the 2 modalities.2-4, 6-11 However, colonoscopic polypectomy and surgical resection have not been compared in a population-based analysis.

We therefore conducted the current study to evaluate patient outcomes of surgical and colonoscopic resection of malignant polyps using the population-based Surveillance Epidemiology and End Results (SEER) database linked to Medicare claims.12, 13 We hypothesized that in selected patients with malignant polyps, polypectomy is associated with favorable outcomes, including a low incidence of subsequent surgery and cancer recurrence. In addition, we hypothesized that the long-term adjusted survival after colonoscopic removal of malignant polyps compares favorably with surgical resection.



The study sample was obtained from patients in the linked SEER-Medicare database who were diagnosed with colon cancer contained within a polyp between 1992 and 2005, with follow-up through 2007. In brief, the database consists of all Medicare-eligible patients who are diagnosed with cancer and reside in any of the population-based registries that are part of the SEER Program.12, 13 Within each registry, approximately 93% of patients aged >64 years are included, with the 7% not captured including those not enrolled in Medicare (3%) or patients who receive care exclusively from the Department of Veterans Affairs. The dataset includes SEER registry data, Medicare enrollment and eligibility criteria, and census tract or ZIP code measures of socioeconomic status, all of which are contained in the Patient Entitlement and Diagnosis Summary Files (PEDSFs). Medicare claims data consist of the Medicare Provider Analysis and Review (MEDPAR) files, which include all hospitalizations, short stay and skilled nursing facility claims; the outpatient files, which include claims from institutional outpatient providers; and the National Claims History files, which include claims from physicians as well as physician assistants, nurse practitioners, independent clinical laboratories, and free-standing ambulatory surgery centers. The files include diagnoses coded by the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) and procedures coded by the Current Procedural Terminology, 4th Edition (CPT-4) or ICD-9-CM. The study also used the 2002 and 2005 Area Resource Files, which contain county-specific measures of physician supply, obtained from the American Medical Association master file.

From an initial sample of 299,260 patients, the following exclusion criteria applied: Medicare eligibility because of end stage renal disease or disability (n = 21,268), previous cancer diagnosis (n = 48,593), histology other than invasive adenocarcinoma (n = 22,087), diagnosis before 1992 or after 2005 (n = 29,174), age at diagnosis <66 years (n = 27,272), cancer other than T1N0M0 within a polyp (n = 145,225), diagnosis at autopsy or from death certificate (n = 3), membership in a Medicare-sponsored HMO or lack of enrollment in Medicare Parts A and B from 1 year before diagnosis through follow-up (n = 1,814), cancer site in the rectum (n = 725), and lack of treatment with surgical resection or polypectomy (n = 22). Thus, the final sample consisted of 2077 patients aged ≥66 years with adenocarcinoma in a colonic polyp who were diagnosed between 1992 and 2005. The sample included 1340 (64.5%) patients treated with surgical resection and 737 (35.5%) patients treated with colonoscopic polypectomy. We excluded patients with malignant rectal polyps from the analysis because surgical options are often less invasive than colectomy (eg, transanal excision).


Data were obtained from the SEER tumor registry files and corresponding Medicare claims. Variables from SEER included demographic characteristics (age, sex, race, marital status, geographic region of residence, and urban/rural status), and tumor characteristics (year of diagnosis, histologic grade, and location in the colon divided into left colon (sigmoid colon to splenic flexure) and right colon (transverse colon to cecum). The SEER Registries were divided into Midwest (Detroit, Iowa), Northeast (Connecticut, New Jersey), South (Atlanta, rural Georgia, Kentucky, Louisiana), and West (San Francisco, Hawaii, New Mexico, Seattle, Utah, San Jose, Los Angeles, Greater California). Small area measures of socioeconomic status included median household income and proportion of adults with at least a high school education. Both measures were at the census tract level.

From the linked Medicare claims, we obtained data from at least 1 year prior to cancer diagnosis and in the follow-up period through calendar year 2007. The index colonoscopy that led to the cancer diagnosis was identified by searching the relevant Medicare files (MEDPAR, physician/supplier, outpatient) for the procedure that most closely corresponded to the cancer diagnosis date in the PEDSFs. Other characteristics from the index colonoscopy included the setting in which the procedure was performed (inpatient, hospital outpatient, ambulatory surgery), and specialty of the endoscopist obtained though an encrypted version of the physician's unique personal identification number. Medicare claims from 12 months through 1 month before cancer diagnosis were also used to obtain a previously validated comorbidity score.14 Other measures obtained from Medicare claims data for any time prior to cancer diagnosis included previous colon resection, prior colonoscopy, a previous diagnosis of colorectal polyps, diagnosis of inflammatory bowel disease, and a family history of gastrointestinal cancer.

The initial cancer treatment method was divided into 2 groups (1) colonoscopic polypectomy (ie, colonoscopy) and (2) surgical, according to procedure codes. Patients were considered to have colonoscopic treatment if the SEER data indicated polypectomy-endoscopic as site-specific surgery and no procedure codes in Medicare claims were present for surgical resection within 6 months (before or after) of the cancer diagnosis date. Patients were categorized as surgical treatment if the SEER data indicated partial colectomy, subtotal colectomy, or total colectomy as site-specific surgery or procedure codes for surgical resection were present within the 6-month period after cancer diagnosis date. The date of the surgical procedure in Medicare claims that corresponded most closely to the month of cancer diagnosis in SEER was considered to be the procedure associated with cancer treatment. The procedure codes for bowel resection included partial colectomy (ICD-9-CM 45.71, 45.72, 45.73, 45.74, 45.75, 45.76, 45.79; CPT-4 44140-44147), and total colectomy (ICD-9-CM 45.8; CPT-4 44155. 44156, 44160).

Outcomes of interest included postprocedure complications, subsequent receipt of surgical procedures, second primary or recurrent colorectal cancer, and long-term survival. For patients treated with colonoscopic resection, the occurrence of postprocedure hemorrhage (ICD-9-CM 998.11) or perforation (ICD-9 569.83, 998.2) requiring hospitalization within 30 days of the procedure was measured using MEDPAR files. For patients treated surgically, Medicare claims data from 1 to 30 days after resection were searched for the performance of a postoperative procedural intervention. As has been suggested,15, 16 this is thought to represent a more complete and valid measure of quality of surgical care than complications in general. Complications and their associated interventions included retrieval of retained foreign body (CPT-4 49085; ICD-9 54.92), management of postoperative shock/hemorrhage (ICD-9 39.98), management of abdominal infection (CPT-4 45020, 45562, 45563, 49020, 49021, 49040, 49041, 49060, 49061, 49080, 49081, 75989; ICD-9 54.0, 54.91, 54.19), repair of an organ injury/laceration (CPT-4 38100, 44602, 44603, 44604, 44605; ICD-9 41.5, 44.61, 46.71, 46.73, 46.75, 48.71, 50.61, 51.91, 69.41), reoperative laparotomy (CPT-4 49002, 49000, 49010; ICD-9 54.12, 54.11, 54.21), management of wound complication (CPT-4 10060, 10061, 10120, 10121, 10140, 10180, 12020, 13160, 97601, 97602; ICD-9 54.61, 86.22), management of fistula (CPT-4 44640, 44650, 44660, 44661, 45800, 45805, 45820, 45825; ICD-9 46.72, 46.74, 46.76), management of bowel obstruction (CPT-4 44050), and management of stoma complication (CPT-4 44340, 44345, 44346; ICD-9 46.40, 46.41, 46.42, 46.43).

For patients in both the colonoscopy and surgical groups, the MEDPAR, outpatient, and physician/supplier files were examined through the end of the follow-up period for claims that indicated surgical resection using the procedure codes defined above. In both groups, occurrence of second primary colorectal cancers was measured through the SEER database. The SEER registry does not measure cancer recurrence. However, using a previously developed algorithm with Medicare data,17 recurrence was determined by the presence of claims for cancer chemotherapy (ICD-9-CM 99.25, V58.11, V67.2, V66.2; CPT-4 Q0083-0085, J9000-J9999; Revenue Center codes 0331, 0332, 0335) or diagnosis codes indicating metastatic disease (lung, 197.0-197.3; liver, 197.7; other sites, 197.4, 197.6, 197.8, 198.0-198.82) from 6 months to 2 years after diagnosis. The sensitivity and specificity of this algorithm for detecting relapsed acute myelogenous leukemia were 86% and 99%, respectively.17 Survival was measured until death or the end of the follow-up period (December 31, 2007).

Statistical Analysis

Descriptive characteristics were compared between the colonoscopy and surgical groups using chi-square analysis and the Wilcoxon rank sum test for categorical and continuous variables, respectively. Survival time was compared between groups using Kaplan-Meier analysis. Because of likely selection bias in treatment allocation, to further examine differences in long-term survival, we used a propensity score adjustment,18 where all variables potentially associated with treatment decisions were included in the multivariable logistic propensity model predicting likelihood of colonoscopic therapy. A propensity score estimates the probability of assignment to a specific treatment given a set of variables. Moreover, because both measured and unmeasured factors are thought to potentially track together, omission of a variable that could not be measured would be expected to be associated with less bias than with traditional regression models. As generally recommended,19 patients were then divided into quintiles of predicted probability of colonoscopic polypectomy, and within each quintile, survival was compared using Cox proportional hazards regression, generating hazard ratios with 95% confidence intervals (95% CI).


Using the specified entry and exclusion criteria, we identified a total of 2077 patients in the SEER-Medicare database, including 1340 (64.5%) treated surgically and 737 (35.5%) treated with colonoscopic polypectomy. Demographic, clinical, and procedural characteristics in both groups are shown in Table 1. The patients treated with polypectomy were older (78.0 vs 74.9 years) and were more likely to be married (P<.001). Polypectomy patients also had higher comorbidity scores and were less likely to have previous colonoscopy or a diagnosis of colonic polyps (P<.001). Tumor characteristics also differed between the 2 groups, with polypectomy less likely to be used in the right colon and more likely to be performed for well-differentiated cancers or those with unknown differentiation (P<.001). There were also specialty differences, with polypectomy more likely to be used in management if the colonoscopy was performed by a gastroenterologist and less likely if it was performed by a general surgeon (P = .002). The frequency of polypectomy increased over the course of the study and was more likely to be used as management in patients who resided in the Northeast and less likely in the Midwest (P = .001).

Table 1. Characteristics in Surgical and Endoscopic Treatment Groups
CharacteristicOverallSurgicalColonoscopic PolypectomyP
  1. Abbreviation: SEER, Surveillance Epidemiology and End Results.

  2. Data are presented as no. (%) unless specified otherwise.

Age group, y    
 66-69421 (20.3)313 (23.4)108 (14.7)<.001
 70-74519 (25.0)361 (26.9)158 (21.4) 
 75-79521 (25.1)342 (25.5)179 (24.3) 
 80-84383 (18.5)200 (14.9)183 (24.8) 
 85+233 (11.2)124 (9.3)109 (14.8) 
Sex   .80
 Women1018 (49.0)654 (48.8)364 (49.4) 
 Men1059 (51.0)686 (51.2)373 (50.6) 
Race   .17
 White1700 (81.9)1108 (82.7)592 (80.3) 
 Black156 (7.5)90 (6.7)66 (9.0) 
 Other221 (10.6)142 (10.6)79 (10.7) 
Marital status   <.001
 Married1,120 (53.9)783 (58.4)337 (45.7) 
 Not married821 (39.5)507 (37.8)315 (42.8) 
 Unknown136 (6.6)51 (3.8)85 (1.5) 
Census tract income, median (range)$41,148 ($31,186-54,757)$40,750 ($30,925-$54,751)$41,869 ($31,648-$54,824).60
Census tract high school education, %, median (range)82.9 (72.8-90.0)83.4 (74.3-90.5)81.8 (70.5-89.5).01
Year of diagnosis   <.001
 1992-1997679 (32.7)482 (35.9)198 (26.9) 
 1998-2001708 (34.1)460 (34.3)248 (33.6) 
 2002-2004690 (33.2)399 (29.8)291 (39.5) 
SEER region   .001
 Midwest472 (22.7)338 (25.2)134 (18.2) 
 Northeast425 (20.5)252 (18.8)173 (23.5) 
 South284 (13.7)174 (13.0)110 (14.9) 
 West896 (43.1)576 (43.0)320 (43.4) 
Charlson comorbidity index   <.001
 01306 (62.9)887 (66.2)419 (56.9) 
 1453 (21.8)281 (21.0)172 (23.3) 
 2182 (8.8)101 (7.5)81 (11.0) 
 3+65 (6.5)71 (5.3)65 (8.8) 
Prior colonoscopy   <.001
 No1726 (83.1)1080 (80.6)646 (87.7) 
 Yes351 (16.9)260 (19.4)91 (12.3) 
Family history of colorectal cancer   .60
 No2024 (97.4)1304 (97.3)720 (97.7) 
 Yes53 (2.6)36 (2.7)17 (2.3) 
Prior colorectal polyp   <.001
 No1528 (73.6)946 (70.6)582 (79.0) 
 Yes549 (26.4)394 (29.4)155 (21.0) 
Tumor grade   <.001
 Well-differentiated466 (22.4)284 (21.2)182 (24.7) 
 Moderately differentiated822 (39.6)565 (42.2)257 (34.9) 
 Poorly differentiated88 (4.2)75 (5.6)13 (1.7) 
 Unspecified701 (33.8)4168 (31.0)285 (38.7) 
Cancer site   <.001
 Left colon1560 (75.1)949 (70.8)611 (82.9) 
 Right colon517 (24.9)391 (29.2)126 (17.1) 
Colonoscopy site   <.001
 Inpatient442 (21.3)253 (18.9)189 (25.6) 
 Outpatient1250 (60.2)812 (60.6)438 (59.4) 
 Ambulatory292 (14.0)209 (15.6)83 (11.3) 
 Other or unknown93 (4.5)66 (4.9)27 (3.7) 
Colonoscopy physician specialty   .002
 Colorectal surgeon44 (2.1)27 (2.0)17 (2.3) 
 General surgeon97 (4.7)80 (6.0)17 (2.3) 
 Gastroenterologist786 (37.8)480 (35.8)306 (41.5) 
 Internal medicine272 (13.1)175 (13.1)97 (13.2) 
 Family practice175 (8.4)117 (8.7)58 (7.9) 
 Other327 (15.8)223 (16.6)104 (14.1) 
 Unknown376 (18.1)238 (17.8)138 (18.7) 
County-level physician supply (per 100,000)    
 Colorectal surgeon0.31 (0-0.50)0.29 (0-0.50)0.31 (0-0.46).47
 General surgeon11.50 (8.15-14.93)11.50 (7.90-14.93)11.50 (8.98-15.34).07
 Gastroenterologist3.69 (2.05-5.73)3.57 (2.05-5.65)3.69 (2.05-5.83).11
 Primary care26.47 (17.69-35.33)26.53 (17.69-36.11)26.18 (17.84-30.69).02
 Primary care/all physicians0.10 (0.07-0.18)0.10 (0.07-0.19)0.10 (0.07-0.16).02

Outcomes were compared between the 2 groups. Postprocedure complications were documented in 6.2% of patients, including 8.8% in the surgical group and 1.4% in the polypectomy group (P<.001). The most frequent complications were reoperative laparotomy (5.2% of surgical patients), wound complications (2.1%), and abdominal infection (1.7%). A second colon resection was performed after 6 months of diagnosis in 7.2% in the surgical group and 4.5% in the polypectomy group underwent surgery after the initial 6-month window (P = .01). Recurrence was documented in Medicare claims in 14.3% of surgical patients and 13.2% of polypectomy patients (P = .49), and second primary colorectal cancers occurred in 4.0% of surgical patients and 3.7% of polypectomy patients (P = .74). Kaplan-Meier survival estimates in the surgical group were 0.92 and 0.75 at 1 year and 5 years, respectively, and were 0.88 and 0.62, respectively, in the polypectomy group (Figure 1).

Figure 1.

Long-term survival among 2077 patients with malignant polyps treated with surgical resection or colonoscopic polypectomy is shown. In unadjusted analysis, survival was improved among patients treated surgically (P<.001 [log-rank test]).

A multivariable logistic regression model was then developed to predict likelihood of polypectomy as management (propensity score) (Table 2). All clinically relevant predictors, regardless of statistical significance, were included in the analysis. However, factors that were significantly associated with polypectomy included older age, higher comorbidity score, more recent date of diagnosis, well-differentiated cancers, and left-sided cancer location. The rates of polypectomy in quintiles 1 though 5 were 15.7%, 24.0%, 33.7%, 41.3%, and 62.7%, respectively.

Table 2. Factors Predicting the Use of Colonoscopic Polypectomy in Multivariable Analysis
CharacteristicMultivariable OR (95% CI)P
  1. Abbreviation: CI, confidence interval; OR, odds ratio; SEER, Surveillance Epidemiology and End Results.

  2. Data are presented as no. (%) unless specified otherwise.

Age group, y  
 66-69 (reference)1
 70-741.30 (0.96-1.76).09
 75-791.59 (1.17-2.15).003
 80-842.79 (2.02-3.84)<.001
 85+2.57 (1.77-3.73)<.001
 Women (reference)1
 Men1.04 (0.84-1.28).73
 White (reference)1
 Black1.30 (0.89-1.92).18
 Others0.78 (0.55-1.10).16
Marital status  
 Married (reference)1
 Not married1.27 (1.02-1.59).03
 Unknown3.73 (2.50-5.54)<.001
Charlson index  
 0 (reference)1
 11.19 (0.94-1.52).15
 21.77 (1.25-2.49).001
 3+1.96 (1.32-2.90)<.001
Census tract median income1.05 (0.99-1.12).09
Census tract high school education0.99 (0.98-1.00).003
Year of diagnosis  
 1992-1997 (reference)1
 1998-20011.26 (0.98-1.62).07
 2002-20041.64 (1.27-2.12)>.001
Residential area  
 Metropolitan (reference)1
 Urban0.99 (0.74-1.33).93
 Rural0.43 (0.17-1.08).07
SEER region  
 Midwest (reference)1
 Northeast1.28 (0.93-1.74).13
 South1.14 (0.80-1.61).47
 West1.39 (1.06-1.83).02
Tumor grade  
 Well-differentiated (reference)1
 Moderately differentiated0.70 (0.55-0.91).01
 Poorly differentiated0.29 (0.15-0.56)<.001
 Unspecified1.10 (0.85-1.42).46
Cancer site  
 Left colon (reference)1
 Right colon0.45 (0.35-0.58)<.001
Prior colon surgery  
 No (reference)1
 Yes0.44 (0.12-1.62).22
Prior colonoscopy  
 No (reference)1
 Yes0.74 (0.51-1.07).11
Prior colorectal polyp  
 No (reference)1
 Yes0.75 (0.56-1.00).05
Personal history of inflammatory bowel disease  
 No (reference)1
 Yes0.71 (0.29-1.74).45
Family history of colorectal cancer  
 No (reference)1
 Yes1.00 (0.53-1.87).99

Using Cox proportional hazards analysis, the unadjusted hazard ratio for death comparing polypectomy with surgery was 1.51 (95% CI 1.31-1.74; P<.001). After adjusting for propensity quintile, this was reduced to 1.15 and did not reach statistical significance (95% CI 0.98-1.33; P = .08). Within each propensity quintile, the risk of death was similar between the 2 groups (interaction test P = .96). Kaplan-Meier plots within each propensity quintile are also shown graphically (Figure 2) and demonstrate no difference in survival between the 2 treatment groups. Postsurgical complication rates were 8.0%, 7.6%, 10.1%, 10.7%, and 7.7% in quintiles 1 through 5, respectively, whereas postpolypectomy complication rates were 1.5%, 0%, 0%, 1.8%, and 1.5%, respectively.

Figure 2.

Long-term survival in patients with malignant polyps treated with surgical resection (solid lines) or colonoscopic polypectomy (dashed lines), stratified by propensity quintile, is shown. Data represent the quintile least likely to undergo polypectomy (A) through the most likely (E). Within each quintile, survival did not differ between groups (quintile 1, P = .21; quintile 2, P = .50; quintile 3, P = .41; quintile 4, P = .71; quintile 5, P = .33 quintile 5 [log-rank test]).


The management of malignant polyps, which are typically grossly benign-appearing lesions that contain invasive carcinoma in the resected specimen, includes colonoscopic and surgical approaches. Clinical algorithms have been developed to help guide subsequent management,1-5 with clinical criteria favoring polypectomy as sole management, including well or moderately differentiated lesions, complete resection (typically absence of malignant cells 1-2 mm from the transected margin), no lymphatic or venous invasion, and no polypoid carcinoma (no benign tissue in the resected polyp). In addition, polyps that are sessile as opposed to pedunculated may be less amenable to colonoscopic management. Studies conducted before the advent of colonoscopy with polypectomy estimated the overall incidence of nodal metastases to be approximately 10%.6 However, after balancing the morbidity and mortality associated with bowel resection with the risk of subsequent recurrence, many clinicians advocate colonoscopic local excision for locally malignant lesions that have favorable histopathologic features. Studies from individual endoscopists,5, 7 single centers,3, 4, 9-11 or pathology databases8 have reported excellent long-term results in patients treated with polypectomy, often approaching 100% 5-year cancer-free survival rates. However, population-based data about the use and effectiveness of colonoscopy as treatment for a malignant polyp are lacking.

Using the population-based SEER-Medicare database, we found that more than 35% of malignant polyps were managed with colonoscopic resection, with the proportion increasing over the course of the study period. In general, patients treated with polypectomy were presumably higher risk for surgery, as reflected in age and comorbidity scores, and were more likely to have tumor characteristics that were favorable, including well-differentiated cancers. Nonetheless, when adjustment was made for clinical factors that predicted likelihood of colonoscopic therapy, long-term survival did not differ between the 2 groups. These findings suggest that as currently practiced, management with polypectomy is a safe and effective strategy.

The frequency and outcomes of colonoscopic management in this study compare favorably with other published case series. In the most recently published large series,8-11 the frequency of favorable pathology ranged from 26% to 47%, and these patients were typically treated colonoscopically. Although favorable long-term outcomes were more likely in patients with low-risk pathological features, in some series more than 80% of patients with high-risk histology also had good outcomes with polypectomy. The lower complication rate with polypectomy (1.4%) compared with surgical treatment (8.8%) in this older cohort of patients may have been biased by excluding patients who underwent surgery within 6 months of polypectomy. Nonetheless, given these data, polypectomy may be a reasonable management strategy in more patients than previously reported.

The datasets and methodology used in the current study have several advantages over those described in previous work. First, the SEER-Medicare data are population-based and capture patients treated in a wide variety of practice settings. Second, the linkages available with the dataset enable longitudinal analyses, including the use of pre- and postcolonoscopy procedures, cancer diagnoses, surgical intervention, and survival data. In addition, the data include information on other factors that could affect treatment decisions, including comorbidity, sociodemographics, and physician supply. As with other published series on malignant polyps, the data are observational, and patients were not randomized to different treatment strategies; however, the use of propensity scores represents an attempt to adjust for selection bias in treatment allocation. Limitations of the data include absence of certain pathological features, including the margins of excision and lymphovascular invasion. In addition, the use of Medicare-eligible patients limited the study population to those aged ≥66 years; however, this age group has the highest incidence of colorectal cancer and adenomatous polyps20, 21 in the United States. The cohort was also limited to fee-for-service Medicare beneficiaries and thus, treatment in managed care populations was not assessed. However, fee-for-service still accounts for the majority of beneficiaries. Finally, certain measures such as comorbidity, cancer recurrence, and postprocedure complications relied on diagnosis and procedure codes in Medicare claims data. Where possible, we used previously validated measures,14-17 though in the current study, chart validation was not performed. In contrast, tumor-related data from SEER are considered to represent a gold standard.

In conclusion, colonoscopic management of malignant polyps is commonly performed in routine clinical practice. Under usual practice conditions, patients with similar clinical characteristics have comparable outcomes with polypectomy or surgical resection, and polypectomy could be offered to patients who meet appropriate clinical criteria.


Supported by National Institutes of Health Grant R01 CA132862 (to G. S. C.).


The authors made no disclosures.